The Agent Skills Directory

[SAFE]: The skill provides legitimate instructions for technical planning, research, and documentation. It encourages skeptical verification of information and human-in-the-loop checkpoints.
[PROMPT_INJECTION]: The skill possesses a surface for Indirect Prompt Injection as it is designed to ingest untrusted external data such as tickets and research documents.
Ingestion points: Reads files and ticket details in Step 1 and Step 2 of the planning process.
Boundary markers: No explicit delimiter-based isolation or 'ignore instructions' warnings are mentioned for processed data.
Capability inventory: Uses Read, Glob, Grep, and Task tools; can spawn auxiliary research agents like codebase-analyzer.
Sanitization: No automated sanitization of ingested content is performed, though the instructions mandate manual verification of all claims against the source code, which serves as a functional mitigation. (Severity: LOW)
[COMMAND_EXECUTION]: The skill uses standard search and read tools to analyze the codebase. There are no patterns suggesting arbitrary shell command execution or unsafe interpolation of user input into system commands.

implementation-planning